Closed-Loop Approach for Robot Dynamic Parameter Identification

Most offline methods for identifying robot dynamics are based on the Inverse Dynamic Model, which is direct in relation to the active parameters. The model is tested as the robot follows reference trajectories that excite the system dynamics, enabling the use of Linear Lest Squares techniques for parameter estimation. The effectiveness of this method has been demonstrated through the experimental identification of numerous examples and industrial robots. Nevertheless, it needs measurements of joint force/torque and position, as well as the estimation of link rate and speeding up, which is achieved by bandpass filtering the link position at high sampling rates. The newly suggested method, called Direct and Inverse Dynamic Identification Model, requires only joint force/torque measurements, eliminating the need to calculate rapidity and acceleration through bandpass sifting of the measured position. This is a closed-loop output error method, where the conventional joint position output is replaced by link force/torque. The method relies on a closed-loop simulation of the robot using the direct dynamic model, applying the same control law construction and reference trajectory for both the real and simulated robots. The optimal parameters are determined by minimizing the 2-norm of the error in the middle of the actual and simulated force/torque values. An authentication experiment conducted on a 2 DOF direct drive rigid robot demonstrates the method's effectiveness.